2,359 Photos in 3.9 Seconds
What my home AI agent just taught me about our own photo library — and one beloved location in particular.
Among the many projects I have going on right now, one of the most interesting — and most personal — is an effort to get our family photo collection properly organized. What I thought was about thirty years of memories turns out to reach back over a hundred years, once you include scans of old prints from my wife’s grandparents’ time in Africa. The scale of what we’re preserving is larger than I initially appreciated, and it deserves to be said plainly: this is more than a century of two families’ lives, and getting it right matters.
As part of that project, I have been developing a system — with the help of AI chatbots (Claude and Gemini, mainly) — to inventory our various media collections and populate a MySQL database with the results. GPS coordinates, keywords, file hashes, metadata — it is all going in. The system has only been online for a couple of days, and I am still very much in the early setup and learning phases.
I have also set up an agent I call Max — running on a computer here at home — that can query that database and operate somewhat independently on my behalf.
A Question at the Pizza Place
In a recent conversation about this project, someone raised an idea that immediately piqued my interest: the ability to query a database for all photos from a specific event or location. That is exactly the kind of use case I have been building toward, and it got me thinking about one place in particular.
My wife Annie and I have a long connection to the Paris France Temple. We visited the site shortly after it was announced. We served there at the Visitors’ Center in 2018. And we go back every year. As you might imagine, we have a lot of photos.
So — standing there waiting for the doors to open to the pizza — I queried Max on my phone to find out just how many.
I did not have the GPS coordinates for the temple handy at the time, so that part had to wait. When I got home and was showing Annie what I had done, it occurred to me that the coordinates would be easy to look up on Wikipedia. I did, provided them to Max, and got my answer:
There are 2,359 photos of the Paris Temple in the database.
I was curious how long the query had taken, so I asked Max directly. His answer:
“Once you provided the GPS coordinates for the Paris Temple, it took me approximately 3.9 seconds to tell you that there are 2,359 photos.”
Fun, no? 😊
You can read the full transcript of the Max interaction here.
Meet the Team
I should briefly introduce the collaborators on this project, because they are not human — which is worth explaining to anyone who hasn’t worked this way before.
Claude is an AI assistant made by Anthropic. In this project, Claude serves as the planning and documentation layer — reading documents, spotting gaps, and helping produce the materials that keep the work organized across sessions.
Max is a separate AI agent running on a computer here at home. Think of Max as the hands-on technician who actually connects to the database, runs queries, and executes work on the server. Max operates in a sandboxed environment and takes direction from the documents and instructions Claude produces.
The division of labor is straightforward: Claude thinks, plans, and documents. Max executes. Mike approves anything that could cause permanent changes.
What About All Those Duplicates?
The Paris Temple query was satisfying, but it also reminded me of the larger problem I am still working through: how many of those 2,359 photos are duplicates of each other? Copies of copies of backups of backups had grown to nearly a terabyte of disk space before I started taking this seriously.
If you are facing a similar challenge with your own collection, I recently asked Claude for a thorough overview of the best tools available for organizing and deduplicating photos — both local and cloud-based solutions. You can read that full response here.
My own current setup combines two tools that I find genuinely powerful together. Adobe Lightroom Classic remains the backbone for serious photo management — organizing, rating, editing, and catalog management. For anyone invested in photography at any level, it is hard to beat.
Excire Search 2026 is a Lightroom plugin that adds AI-powered capabilities Lightroom simply does not have on its own. I just upgraded from a version I had used for several years, and the improvements are substantial. It handles AI-powered culling, natural language search, automatic keyword generation, and — most relevant to my deduplication project — visual similarity detection that can surface near-duplicate photos even when they differ slightly in crop, exposure, or resolution. Everything runs locally; your photos never leave your computer.
Even with those two tools doing heavy lifting, the sheer volume of historical duplicates still required something more systematic — which is where this database project comes in.
What Comes Next
The database is young. Max is young. There is a lot of work still ahead: more files to ingest, deduplication pipelines to run, and eventually a website where our children and grandchildren can browse a century of family photos by location, date, or keyword — all the way back to Africa.
But 2,359 photos of one beloved location, surfaced in 3.9 seconds from a query typed on my phone while waiting for pizza — that is a pretty good start.
“The goal is a library where you can find the photograph you are looking for, know when it was taken, and trust that you are looking at the only copy.”
This post reflects work in progress as of March 2026. Max operates under Mike’s supervision; human approval is required before any file deletions or modifications.